ORIGINAL PAPER

A definition of mountains and their bioclimatic belts for globalcomparisons of biodiversity data

Christian Korner • Jens Paulsen • Eva M. Spehn

Received: 9 December 2010 / Revised: 8 July 2011 / Accepted: 19 July 2011 / Published online: 25 August 2011

� The Author(s) 2011. This article is published with open access at Springerlink.com

Abstract This is the first quantitative attempt at a global

areal definition of ‘alpine’ and ‘montane’ terrain by com-

bining geographical information systems for topography

with bioclimatic criteria (temperature) subdividing the life

zones along elevational gradients. The mountain definition

adopted here refrains from any truncation by low elevation

thresholds, and defines the world’s mountains by a com-

mon ruggedness threshold ([200 m difference in elevation

within a 2.50 cell, 0.50 resolution), arriving at 16.5 Mio km2

or 12.3% of all terrestrial land area outside Antartica being

mountains. The model employed accounts for criteria of

‘‘mountainous terrain’’ for biological analysis, and thus

arrives at a smaller land area fraction than hydrologically

oriented approaches, and by its 2.50 resolution, it includes

less unstructured terrain (such as large plateaus, very wide

valleys or basins) than earlier approaches. The thermal

delineation of the alpine and nival biogeographic region by

the climatic tree limit (the lower boundary of the alpine

belt) arrives at 2.6% or 3.55 Mio km2 of the global land

area outside Antarctica (21.5% of all mountain terrain).

Seven climate-defined life zones in mountains facilitate

large-scale (global) comparisons of biodiversity informa-

tion as used in the new electronic ‘Mountain Biodiversity

Portal’ of the Global Mountain Biodiversity Assessment

(GMBA).

Keywords Mountain definition � Ruggedness �Mountain life zones � Alpine � Montane �Thermal belts

Introduction

Much of the Earth’s terrestrial surface is covered by moun-

tains, which host a larger proportion of the Earth’s biodiversity

than would be expected by area (Korner 2004, Mutke and

Barthlott 2005). Due to steep environmental gradients over

short distances, mountains exemplify ‘natural experiments’

that permit testing ecological theories and questions of

adaptive evolution (Korner 2000, Korner et al. 2007).

In recent years, legacy data on species’ distributions

(most often hosted in museums and natural history collec-

tions) have become available in digital form. To the extent

such data are geo-referenced, including precise information

on elevation, they can be linked to geographical information

systems on topography, climate, geology, etc. Such elec-

tronic archives offer a new way to explore biodiversity, its

causes, and evolution (Korner et al. 2007, Spehn and Korner

2010). The electronic ‘Mountain Biodiversity Portal’ of the

Global Mountain Biodiversity Assessment of DIVERSITAS

(GMBA 2010) is a thematic portal for mountains with open

access to biological data hosted by the Global Biodiversity

Information Facility (GBIF). The user is offered a mountain

relevant area-selection, with a horizontal (region) and ver-

tical (elevation, climate) dimension, based on a coherent

convention of terms. The Mountain Biodiversity Portal aims

at becoming a standard tool for the world community of

mountain biologists and ecologists. This article presents the

conceptual framework of this biogeographical mountain

convention.

GMBA definition of mountains

While seemingly obvious to most people, it is very difficult

to offer a quantitative generalizable scientific definition of

C. Korner � J. Paulsen � E. M. Spehn (&)

Institute of Botany, University of Basel,

Schoenbeinstrasse 6, 4056 Basel, Switzerland

e-mail: gmba@unibas.ch

Alp Botany (2011) 121:73–78

DOI 10.1007/s00035-011-0094-4

123

what a mountain is that can be used for mountain-specific

data retrieval in biodiversity research. A mountain cannot be

defined by elevation, simply because there are elevated

plateaus such as the North-American short-grass Prairies at

around 2,000 m elevation or the vast plateaus in central

Asia, while steep coastal ranges may exemplify ‘real’

mountains near sea level. Similarly, mountains cannot be

defined by climate, given that any cold category would

include arctic and antarctic lowland, and tropical mountains

range from equatorial rain forests to arctic life conditions

near their summits.

The only common feature of mountains is their steepness

(slope angle to the horizontal) which causes the forces of

gravity to shape them and create habitat types and distur-

bances typical for mountains and which make exposure a

driving factor of life (Korner 2004). Because steepness is a

feature of each specific slope that cannot be quantified at a

spatial scale of such a global database, the Mountain Bio-

diversity Portal adopts ruggedness as a simple and pragmatic

proxy for steepness, to define mountains across the globe.

Ruggedness is defined here as the maximal elevational

difference among neighbouring grid points. Calculations are

based on the digital elevation model (DEM) used by

WorldClim (Hijmans et al. 2005). Elevation of every cell in

a 3000 grid was compared with elevation of its eight neigh-

boring cells. If the difference between the lowest and highest

of these nine 3000 grid cells exceeds 200 m, the central cell is

assigned as ‘rugged’ i.e. belonging to mountain terrain, as a

matter of convention. We then reduced the dataset to a

resolution of 203000 (by using every 5th 3000 cell in latitude

and longitude) for the final calculation of ruggedness, to

arrive at a manageable dataset, mainly because WorldClim

Climate data are on a 203000 grid only. The Mountain Bio-

diversity Portal therefore operates at a 20 3000 resolution of

the terrestrial surface (c. 4.6 9 4.6 km or 21.5 km2 at the

equator and narrower at higher latitudes, i.e. cos �latitude,

resulting in 15.2 km2 at 45�, and 10.7 km2 at 60� latitude).

We gave the selection of the ruggedness threshold of

200 m elevational difference (in 9 3000cells) a lot of thought.

In a logarithmic land area versus ruggedness plot, a 200 m

ruggedness threshold roughly marks the point at which the

land area starts to decline logarithmically as ruggedness is

further increased. A main criterion with regard to biodi-

versity was that the threshold is inclusive rather than

exclusive with regard to valley floors, adjacent forelands and

plateaus. The 200 m threshold turned out to meet this

demand well (Fig. 1). As can be seen in this example for the

transition of the Swiss Alps to the Swiss midlands (or Swiss

plateau), the model includes all mountain valleys except the

very widest (approx. [2.5 km width). The patterns around

Lake Brienz (the top right lake in Fig. 1) shows that

mountain pixels extend substantially into flat terrain and

foothills. A few hills in the otherwise even lowlands (mainly

agricultural land between 300 and 500 m elevation) are also

depicted as ‘mountains’. We intentionally used a reference

map with roads and cities in Fig. 1 to visualize the mountain-

lowland contrast. Earlier mountain definitions (such as

Meybeck et al. 2001, and others, see below) would place

most of that hilly lowland terrain into the mountain category,

which might make sense e.g. in a hydrological context, but

would seem inappropriate in a mountain biodiversity con-

text. With this definition, 16.5 Mio km2 or 12.3 % of the

terrestrial surface is rugged at this scale (Table 1 offers

results for 3 different ruggedness thresholds).

Ruggedness, as defined here, may refer to a single

[200 m elevational distance between two out of nine

neighboring 3000 grid cells on a 20 3000 scale. The remainder

of its area can exhibit low inclination terrain (valley floors,

small plateaus, forelands), therefore this convention also

covers non-rugged terrain adjacent to mountains at the given

20 3000 resolution (4.6 9 4.6 km at the equator). In rare cases

a pixel may not be assigned rugged, although it is, because

the grid failed to capture a certain landscape feature at the

3000 (0.9 km) resolution. Needless to say that no topographic

information\3000 is reflected in these data, hence, also the

boundary of mountains adjacent to lowland is not more

accurate than 3000 (c. 0.9 km at the equator).

Earlier attempts to define mountains go back to the 19th

century, and used several criteria such as elevation, volume,

relief and steepness, but have been inconsistent on a global

scale (Gerrard 1990). A recent attempt to arrive at a global

mountain convention by Kapos et al. 2000 used a mixture of

elevation and ruggedness criteria (elevation [2.500 m; or

1,500–2,499 m if the slope is[2�; or 1,000–1,499 m if the

slope is 5� and the local elevation range at a radius of 7 km is

[300 m; or 300–999 m if the local elevation range at a

radius 7 km is[300 m). Meybeck et al. 2001 used basically

the same approach and resolution as we did, defining

mountains with a fixed relief roughness at a resolution of

300 9 300 (RR = maximum minus minimum elevation per

cell divided by half the cell width). The main difference is

that Meybeck et al. 2001 used a cut off towards the lower end

of 500 m elevation and used an even lower ruggedness

threshold (40%), whereas we use ruggedness with a single,

higher threshold (200 m or 77%) only, independent of

meters of elevation. Both these earlier definitions that used a

cut-off elevation (300 or 500 m) had been selected for

hydrological (Meybeck et al. 2001) or mountain forest

questions (Kapos et al. 2000). For mountain biodiversity, it

seems to be appropriate to restrict forelands and valleys to

\2 km distance to mountains thus including the immediate

forelands or plateaus to that extent only.

By including less structured terrain, both earlier defini-

tions arrived at a larger extent of mountain terrain (20.9% of

total land area in Meybeck et al. 2001 at their[40 m km-1

category), and 24% in Kapos et al. 2000. As discussed by

74 Alp Botany (2011) 121:73–78

123

Meybeck et al. 2001, the larger area of Kapos et al. 2000 is

most likely due to the extensive inclusion of high plateaus

such as the Tibetan Plateau, and a coarser scale and thus

greater tolerance of including forelands and basins (Kapos

et al. 2000). Meybeck et al. 2001 adapted a ‘degree of

dissectedness’ of 40–80 m km-1 (40–80%) as moderately

dissected terrain that may be seen as separating hills from

mountains. Our 200 m ruggedness threshold corresponds to

77% degree of dissectedness of Meybeck et al. 2001. As our

analysis (Table 1) shows, a ruggedness of only 50 m instead

Fig. 1 Top Mountain area according to our ruggedness definition

(elevational distance [200 m between nine 3000 pixels) on the global

scale (black, non-rugged terrain in white). Below Mountain area on

the regional scale (blue, non-rugged terrain in yellow). The below

map shows a part of Switzerland from Neuchatel to Grindelwald

(6.78075 E, 47.03915 N, 8.21894 W, 46.40867 S). Note the encircled

area for lake Brienz illustrating the extent of flat mountain foreland

terrain included in our mountain definition. All mountain valleys

except small parts of the very widest valleys are covered by our

mountain definition. Topographic map by [http://map.geo.admin.ch]

Alp Botany (2011) 121:73–78 75

123

of our 200 m threshold (i.e. a single elevational contrast of

50 m in a 4.6 9 4.6 km area near the equator; equal to ca.

10–20 m km-1 in Meybeck et al. 2001 attributed by them as

nearly flat) almost doubles the mountain area.

Seven thermal belts

For a global comparison of mountain biota it is essential that

the latitudinal change in life conditions with elevation is

accounted for. Hence, elevational belts have to be converted

into climatic belts that account for the rise of isotherms as

one approaches the equator. Here we suggest subdividing

mountains vertically into seven thermal life zones (thermal

belts) defined by temperature only and, thus, accounting for

the latitudinal change in elevation of thermally similar areas.

All belts refer to the best defined biome boundary in

mountains, the high elevation climatic treeline, separating

the treeless alpine and the potentially forested montane

belts. From there, one can go up (alpine and nival) or down

(montane and lower) based on temperature criteria

(Table 2). Thermal belts are defined by a model using

WorldClim climate data (daily air temperatures and snow

cover) and field data from across the globe that characterize

the position of the potential high elevation climatic treeline,

irrespective of the actual presence or absence of trees in a

given area (Korner and Paulsen 2004 and additional data)

High elevation treeline

The Mountain Biodiversity Portal adopts the position of

the potential, climatic high elevation treeline as the main

reference line for life zones in mountains. Defined by an

isotherm, it exerts an ideal bioclimatic reference line for

any comparison of mountain biota worldwide (Korner

2007). The treeline may be located at a few hundred

meters above sea level near the arctic circle, but may

reach [4,000 m in the tropics and subtropics (as long as

annual precipitation is [250 mm). The climatic treeline

marks the limit of tall upright life forms that are aero-

dynamically strongly coupled to the free atmosphere and,

thus, are facing thermal constraints well represented by

weather station data (Korner 2007). In contrast, low

stature shrub- or grass-type vegetation at least periodically

decouples aerodynamically from ambient conditions and

experiences/produces peculiar microclimates, substantially

warmer than what climate stations would report (Scherrer

and Korner 2010). The transition from potentially forested

to treeless terrain is co-defined by an empirically deter-

mined minimum duration of the growing season of

94 days and a mean growing season temperature of 6.4�C.

Where trees or any other vegetation is naturally absent

e.g. due to lack of moisture, this line is still used as an

isotherm that separates terrain above and below (hence,

there may be alpine deserts and montane deserts). The

Table 1 Terrestrial land area outside Antarctica (a total of 134.6 Mio km2) subdivided by different thresholds of ruggedness (R) expressed as

maximum contrast in elevation among 9 pixels of 3000 in each cell of a 203000geographical grid

R (meters of elevation) Continent/region (Mio km2)

As Eu Af N-A S-A Gld Aus Oce Total %

All 44.6 9.8 30.0 22.1 17.8 2.1 7.7 0.5 134.6 100.0

\50 23.5 6.7 23.5 14.0 11.8 1.8 6.8 0.1 88.2 65.5

C50 \200 12.2 2.2 5.3 5.2 3.8 0.2 0.8 0.2 9.9 22.2

C200 8.9 0.9 1.2 2.9 2.2 0.1 0.1 0.2 16.5 12.3

As Asia, Ee Europe, Af Africa, N-A North America, S-A South America, Gld greenland, Aus Australia and New Zealand, Oce Oceania (including

the large islands of SE Asia)

Table 2 The global area of

bioclimatic mountain belts for

rugged terrain

Temperatures refer to growing

season (GS) mean temperatures

(for definition see text). M,

percentage of total mountain

area (100% = 16.5 Mio km2),

G, percentage of total terrestrial

area outside Antarctica

(100% = 134.6 Mio km2)

Thermal belts Area (Mio km2 ) M (%) G (%)

1. Nival (\3.5�C, GS \ 10 days) 0.53 3.24 0.40

2. Upper alpine (\3.5 �C, GS [ 10 days \ 54 days) 0.75 4.53 0.56

3. Lower alpine \6.4�C, GS \ 94 days) 2.27 13.74 1.68

The treeline

4. Upper montane ([6.4 B10 �C) 3.39 20.53 2.51

5. Lower montane ([10 B15 �C) 3.74 22.64 2.78

6. Remaining mountain area with frost ([15�C) 1.34 8.11 0.99

7. Remaining mountain area without frost ([15�C) 4.49 27.22 3.34

Total 16.51 100.00 12.26

76 Alp Botany (2011) 121:73–78

123

growing season is defined from smoothed perennial time

series in WorldClim by the first transition of the daily

mean of air temperatures per 2.50 grid through 0.9�C, and

its fall below 0.9�C at the end. These numbers have been

obtained from iterative searches for best parameterization

of the model across several hundred reference points

across the globe, improving the criteria as originally

presented by Korner and Paulsen 2004.

Additional isotherms

Below the treeline isotherm, season length may be anywhere

between 95 and 365 days. A further critical, biologically

relevant threshold is the occurrence of freezing. Hence the

lowest mountain belt is defined by the complete absence of

freezing (the ‘banana’ belt). Since WorldClim does not offer

absolute minima of temperature at hourly resolution,

Mou

ntai

n la

nd a

rea

per

ther

mal

bel

t (M

io k

m2)

-60 -50 -40 -30 -20 -10 0 10 20 30 40 50 60 70 800

0.3

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

warm, no freezingwarm, freezinglower montaneupper montane

lower alpineupper alpinenival

Latitude (°)South North

thermal treeline

Thermal belts of the world‘s mountainsFig. 2 Mountain land area

(Mio km2) for each thermal belt

across all latitudes. Note the

Northern versus Southern

Hemisphere asymmetry of

mountain land area. About one

third of all mountain terrain is

presumably frost free (orange)

Fig. 3 Thermal belts of Hawaii’ Big Island (19.583333�, -155.5�)

mountains (Mauna Kea, 4205 m a.s.l in the north, Mauna Loa 4169 m

a.s.l in the south). See Fig. 2 for colour codes of thermal belts. Satellite

picture by Jacques Descloitres [http://commons.wikimedia.org/

wiki/File:Hawaje.jpg]

Alp Botany (2011) 121:73–78 77

123

absence of freezing was assumed, when the lowest daily

mean temperature was [5�C. This assumption is likely to

underestimate the likelihood of climatic extreme events that

may be biologically decisive in open terrain. Hence the

completely freezing-free area is possibly smaller than

assumed here. The thermal belts as defined here cover all

possible moisture regimes from permanently humid to arid

and are thus, not specific to a certain type of vegetation.

Therefore, these thermal belts make the mountains of the

world comparable across latitudes, irrespective of their

elevation in meters and moisture regimes (Table 2). With

the model parameters as defined above, season length and

season mean temperature were calculated for all

WORLDCLIM cells (on a 203000 grid), allowing to assign

each cell to one of our thermal belt classes (Table 2). The

total area of each belt was then calculated as the number of

grid cells multiplied by the cell area. (Figs. 2, 3)

A note of caution: working with gridded data, the accu-

racy of an analysis usually increases with the number of grid

cells included, depending spatial resolution and extent of the

phenomenon being mapped. Hence, best results are obtained

for large areas (e.g. Alps, Pyrenees, Cascades, Hindu-Kush

Himalaya, etc.) across which local statistical deviations

from reality of both topography criteria (ruggedness) and

climate data become less significant. A single grid cell may

deviate from the nearest grid cell by kilometers of elevation

in the case of steep mountain flanks. Further, the map is a

Mercator projection and, thus, the spatial size of grid cells

depends on latitude.

In summary, using these definitions, the global land area

above the treeline isotherm comprises 3.55 Mio km2 or

21.51 % of all mountain terrain (or 2.64 % of all land outside

Antarctica). Twenty seven percent of all mountain terrain

falls in the warm, low elevation category. This surprisingly

large rugged area represents the lower slopes and foothills of

warm temperate, subtropical and tropical mountains.

We advise against the use of elevation in meters when

defining a lower and an upper limit of biota across larger

areas. Even in regional studies, there is a risk of climatic

bias, because, for instance, front ranges and central ranges

may differ dramatically in climate. Since life in mountains is

not driven by elevation per se, but by the climatic conditions

associated with elevation, thermal belts of life offer a simple,

temperature-only driven zonation of mountains.

This mountain convention offers a means for consistent

comparison of mountain life zones at global scales, and we

hope that a large body of scientific works will emerge from

the Mountain Biodiversity Portal.

Acknowledgments We thank Falk Huettmann for comments on an

earlier draft of the manuscript, and three reviewers for their con-

structive comments. The Mountain Biodiversity Portal (http://www.

mountainbiodiversity.org) is a project by the Global Mountain Bio-

diversity Assessment (GMBA, Basel, Switzerland) of DIVERSITAS

(Paris), in cooperation with the Global Biodiversity Information

Facility (GBIF, Copenhagen, Denmark) and is funded by the Swiss

National Science Foundation (31FI3A-118167 to Ch. Korner).

Open Access This article is distributed under the terms of the

Creative Commons Attribution Noncommercial License which per-

mits any noncommercial use, distribution, and reproduction in any

medium, provided the original author(s) and source are credited.

References

Gerrard J (1990) Mountain environments: an examination of the

physical geography of mountains. Belhaven Press, London

GMBA (2010) The digital ‘Mountain Biodiversity Portal’.

http://www.mountainbiodiversity.org

Hijmans RJ, Cameron SE, Parra JL, Jones PG, Jarvis A (2005) Very

high resolution interpolated climate surfaces for global land

areas. Int J Climatol 25:1965–1978

Kapos V, Rhind J, Edwards M, Price MF, Ravilious C (2000)

Developing a map of the world’s mountain forests. In: Price MF,

Butt N (eds) Forests in sustainable mountain development

(IUFRO Research Series 5). CABI Publishing, Wallingford

Oxon, pp 4–9

Korner Ch (2000) Why are there global gradients in species richness?

Mountains may hold the answer. TREE 15:513

Korner Ch (2004) Mountain biodiversity, its causes and function.

Ambio Special Rep 13:11–17

Korner Ch (2007) Climatic treelines: conventions, global patterns,

causes. Erdkunde 61:316–324

Korner Ch, Paulsen J (2004) A world-wide study of high altitude

treeline temperatures. J Biogeogr 31:713–732

Korner Ch, Donoghue M, Fabbro T, Hauser C, Nogues-Bravo D,

Arroyo MTK, Soberon J, Speers L, Spehn E, Sun H, Tribsch A,

Tykarski P, Zbinden N (2007) Creative use of mountain

biodiversity databases: The Kazbegi research agenda of

GMBA-DIVERSITAS. Mt Res Dev 27:276–281

Meybeck M, Green P, Vorosmarty CJ (2001) A new typology for

mountains and other relief classes: An application to global

continental water resources and population distribution. Mt Res

Dev 21:34–45

Mutke J, Barthlott W (2005) Patterns of vascular plant diversity at

continental to global scales. Biologiske Skrifter 55:521–531

Scherrer D, Korner Ch (2010) Infra-red thermometry of alpine

landscapes challenges climatic warming projections. Glob

Change Biol 16:2602–2613

Spehn EM, Korner Ch (2010) Data mining for global trends in

mountain biodiversity. CRC Press, Taylor and Francis Group,

Boca Raton

78 Alp Botany (2011) 121:73–78

123